The food additive EDTA aggravates colitis and colon carcinogenesis in mouse models

Abstract

Inflammatory bowel disease is a group of conditions with rising incidence caused by genetic and environmental factors including diet. The chelator ethylenediaminetetraacetate (EDTA) is widely used by the food and pharmaceutical industry among numerous other applications, leading to a considerable environmental exposure. Numerous safety studies in healthy animals have revealed no relevant toxicity by EDTA. Here we show that, in the presence of intestinal inflammation, EDTA is surprisingly capable of massively exacerbating inflammation and even inducing colorectal carcinogenesis at doses that are presumed to be safe. This toxicity is evident in two biologically different mouse models of inflammatory bowel disease, the AOM/DSS and the IL10^-/- model. The mechanism of this effect may be attributed to disruption of intercellular contacts as demonstrated by in vivo confocal endomicroscopy, electron microscopy and cell culture studies. Our findings add EDTA to the list of food additives that might be detrimental in the presence of intestinal inflammation, but the toxicity of which may have been missed by regulatory safety testing procedures that utilize only healthy models. We conclude that the current use of EDTA especially in food and pharmaceuticals should be reconsidered. Moreover, we suggest that intestinal inflammatory models should be implemented in the testing of food additives to account for the exposure of this primary organ to environmental and dietary stress.

Figure 1

Fe-EDTA but not other iron compounds increase colitis activity and colorectal carcinogenesis in the AOM/DSS and IL10^−/− models of IBD. (a,b) Time course of the clinical disease activity index DAI in the AOM/DSS (a) and IL10^−/− (b) model. The timeline of experimental interventions is shown on the x axis. The omitted DSS cycles in the Fe-EDTA group due to high colitis activity are marked with an asterisk. (c,d) Mean DAI over the full time course (i.e. weeks 2–9 for AOM/DSS (c), weeks 1–8 for IL10^−/− (d)); (e,f) Histological activity index (HAI) for AOM/DSS (e) or IL10^−/− (f); (g,h) Tumour burden (i.e., total tumour area per mouse) for AOM/DSS (g) or IL10^−/− (h); exemplary image of hematoxylin–eosin-stained intestines of control (i,k) and Fe-EDTA-fed (j,l) animals. (i) DSS-induced increased inflammatory infiltrate (arrow) with partial loss of crypts in a control animal from the AOM/DSS model. (j) massive inflammation (double arrow) with complete crypt destruction and a single regeneratory layer of epithelial cells covering the lamina propria (single arrow) in an Fe-EDTA-treated animal from the AOM/DSS model. On the lower magnification image (left side), an invasive tumour (*) is seen; the point of invasion through the lamina mucularis mucosae is marked with **. (k) Inflammatory infiltrate (double arrow) and cryptitis through invading neutrophils (single arrow) in a control animal from the IL10^−/− model. (l) Marked hyperplasia, crypt abscess (single arrow) and massive inflammatory infiltrate (lymphocyte aggregates; double arrow) in an Fe-EDTA-treated animal from the IL10^−/− model. Error bars represent standard deviations. Asterisks (*: p < 0.05; **: p < 0.01; ***: p < 0.001) denote statistically significant results compared to the control group.

Figure 2

EDTA compounds enhance colitis activity and colorectal carcinogenesis in the AOM/DSS and IL10^−/− models. (a,b) Time course of the clinical disease activity index DAI in the AOM/DSS (a) and IL10^−/− (b) model. The timeline of experimental interventions is shown on the x axis. (c,d) Mean DAI over the full time course ((i.e. weeks 3–9 for AOM/DSS (c) and weeks 2–11 for IL10^−/− (d)); (e,f) Histological activity index (HAI) for AOM/DSS (e) or IL10^−/− (f); (g,h) Tumour burden (i.e., total tumour area per mouse) for AOM/DSS (g) or IL10^−/− (h). Error bars represent standard deviations. Asterisks (*: p < 0.05; **: p < 0.01; ***: p < 0.001) denote statistically significant results compared to the control group; hashtags (#: p < 0.05; ##: p < 0.01; ###: p < 0.001) mark significant differences between both EDTA compound doses.

Figure 3

EDTA compounds increase paracellular permeability by damaging intercellular contacts. (a) Results of the in vitro FITC-dextran permeability assay on T84 cell monolayers exposed to EDTA compounds and/or IFNγ plus TNFα as noted. Error bars represent standard deviations. Asterisks (*: p < 0.05; **: p < 0.01; ***: p < 0.001) denote statistically significant results compared to the control group, hashtags (^#: p < 0.05; ^##: p < 0.01; ^###: p < 0.001) mark significant differences compared to the IFNγ plus TNFα treated monolayers, and double daggers (^‡: p < 0.05; ^‡‡: p < 0.01; ^‡‡‡: p < 0.001) indicate significant comparisons between EDTA compounds with or without IFNγ plus TNFα. (b,c) Confocal laser endomicroscopy of intestinal epithelium in healthy mice pretreated with Na-EDTA rectally (c) or sham treated (controls; b). The arrow shows accumulation of fluorescein in the crypt lumen and paracellular fluorescein plumes with Na-EDTA. (d,e) Transmission electron microscopy of intestinal tissues from healthy mice treated with Na-EDTA rectally (e) or controls (d). The arrow demonstrates gaps in the intercellular space indicative of breakage of intercellular contacts and specifically of AJs by Na-EDTA. The intercellular contacts in control animals are intact.

Figure 4

EDTA compounds induce dysbiosis in both the AOM/DSS and IL10^−/− models. (a,b) Meta-NDMS plots demonstrating significantly different microbial composition in control animals versus EDTA-treated animals in the AOM/DSS (a) and IL10^−/− model (b). (c,d) Shannon index as a measure of microbial diversity in the AOM/DSS (c) and IL10^−/− model (d). (e–j) Relative abundance of the most abundant and differentially represented microbial genera Akkermansia (e,f) and Peptostreptococcaceae (unknown genus) (g,h) in the AOM/DSS (e,g) and IL10^−/− model (f,h). Error bars represent standard deviations. Asterisks (*: p < 0.05; **: p < 0.01; ***: p < 0.001) denote statistically significant results compared to the control group.